Uber Responds To BuzzFeed News Story On Sex Assault Complaints
"We are working hard to ensure passengers everywhere can get a safe, reliable ride, as well as to provide great customer service when things go wrong," notes Uber.
Following a BuzzFeed News investigation into Uber's internal customer service system — which led to Uber's revelation that the company received five claims of rape, and 170 claims of sexual assault directly related to an Uber ride as inbound tickets to its customer service database between December 2012 and August 2015 — Uber responded shortly after publication with the following message, which we are publishing in full.
Dear Ben, John and Mat,
We have worked with your team over the last week to answer questions about safety at Uber, but the story you have just posted fails to take into account the many facts we have provided to Buzzfeed. Given that our business depends on accountability and transparency, we are now sharing that information publicly so that our riders and drivers have the facts and can judge for themselves.
Getting people from point A to B safely and reliably is the single most important thing Uber does. It’s why we have a dedicated Trust and Safety team, overseen by Joe Sullivan (whose entire career has been focused in this field, first as a federal prosecutor and then at eBay, Facebook, and now Uber) and run by Phil Cardenas (the former head of Trust and Safety at Airbnb).
This team exists to reduce safety incidents, and its success is judged on that one metric. Because even one incident is too many. It’s why Uber has invested heavily in technology to improve safety for everyone before, during, and after each ride. Every Uber trip is GPS-tracked and passengers can share their route in real time with family or friends, as well as rate their drivers at the end of each trip (and vice versa). This is on top of a robust system of background checks.
Sadly, no means of transportation is 100 percent safe today. Accidents and incidents do happen. It’s why we are working to build an exceptional customer support team that can handle problems when they occur, including working with law enforcement.
You asked about screenshots in your possession (and since published) showing that if a customer service representative types “rape” or “sexual assault” into our database, they will see more than 5,800 results (i.e. customer support tickets) for rape and 6,160 for sexual assault over a period of three years. These results are highly misleading because:
Riders routinely misspell “rate” (as in the fare) as “rape”, or use the word “rape” in another context. For example, “you raped my wallet”;
Any email address or rider/driver last name that contains the letters R, A, P, E consecutively (for example, Don Draper) are included. After analyzing the data, we found more than 11,000 rider names and 17,500 rider emails with the letters “rape”;
The results also showed tickets from passengers who got into cars not on the Uber platform, or who were discussing unsubstantiated media reports of sexual assaults.
Our analysis for all of these results shows five tickets that allege an actual rape occurred (0.0000009% of rides in the three years from December 2012 to August 2015) and 170 tickets with a legitimate claim of sexual assault (1 in every 3.3 million trips). Bear in mind that when serious incidents occur, people often report them directly to law enforcement. Therefore, those incidents may not be reflected in the numbers above.
When serious incidents are reported to us, we always reach out to the person who filed the report and, where appropriate, engage with law enforcement. We also temporarily suspend the driver or rider (if it is a question of violence by a passenger) during the investigation.
You asked to have one of your reporters sit with Uber’s customer service team so they could review these types of tickets and validate our numbers. It is entirely fair to ask questions of Uber -- it is the purpose of a free press. But it is unfair to suggest that you cannot trust the veracity of the numbers Uber has provided without personally verifying them, which would be a serious breach of our riders’ and drivers’ privacy.
In addition, you have asked if this is the first time we have audited the data behind the screenshots in your possession. Because of the flaws highlighted above, we do manual audits of every ticket sent to Uber, not audits of key words. It’s why it took us more than two days to answer your questions in detail. It is entirely untrue to conclude that we do not audit safety on our platform. We do, regularly. As we said earlier, our Trust and Safety team is measured internally on one metric: whether it has cut safety incidents on our platform.
Finally, you asked yesterday if Uber had contacted customer service representatives who had recently queried the terms “rape” and “sexual assault” in our database. The answer is yes. We are unsurprisingly concerned that sensitive, personal and confidential data has been shared with people outside Uber. We believe that any company in a similar situation would do exactly the same.
Uber is a relatively young company and we’re the first to admit that we haven’t always gotten things right. But we are working hard to ensure passengers everywhere can get a safe, reliable ride, as well as to provide great customer service when things go wrong. It’s a shame that your article does not reflect much of the substance we have provided.
Joe Sullivan, Chief Safety Officer
Jill Hazelbaker, Vice President, Communications & Public Policy
Tim Collins, Vice President, Global Support
On Monday, Uber posted a correction to the above letter:
Zendesk, one of our customer support platforms, contacted us to say that their search tool would not return a name such as “Don Draper” when searching for the word “rape.” However, such a search would (and did) return names that start with the letters R,A,P,E — even if the ticket itself had nothing to do with a claim of rape. We apologize to Zendesk for using an imperfect (and fictitious) example that doesn’t accurately represent their search functionality. This does not impact our analysis of the overall numbers, which was based on a manual review of these tickets rather than a simple keyword search.